The present invention relates to a technology for simultaneously testing a plurality of integrated circuits formed on a semiconductor wafer in a wafer condition.
Advances in size reduction and cost reduction of recent electronic components incorporating semiconductor IC (integrated circuit) devices are so remarkable that requirements to the size reduction and cost reduction of the semiconductor IC devices are becoming more severe.
In general, a manufacturing of a semiconductor IC device is as follows. First of all, a semiconductor chip and a lead frame are electrically connected with each other by means of a bonding wire. Thereafter, the semiconductor chip is sealed by resin or ceramic and then mounted on a printed board. However, the requirements of reducing the size of an electronic component have introduced a method of directly mounting on a circuit board a semiconductor IC device in a bare chip or chip condition to guarantee quality at low cost. The bare chip or chip condition generally represents a condition of semiconductor IC device which is just cut off a semiconductor wafer.
In order to guarantee quality of bare chips, it is necessary to execute a burn-in screening of a semiconductor IC device in the wafer condition.
However, the burn-in screening of the semiconductor wafer is complicated in handling the semiconductor wafer; thus, the requirement to low cost could not be satisfied. Furthermore, executing the burn-in screening of the plural bare chips on a semiconductor wafer is time-consuming since it requires to execute the burn-in screening separately and repeatedly one by ene or group by group. Thus, in view of time and cost, it does not bring practical merits.
Accordingly, it is earnestly required to realize a simultaneous execution of the burn-in screening of all the bare chips in the wafer condition.
FIG. 32 is a schematic view showing a testing method of a semiconductor wafer using a conventional prober. As shown in FIG. 32, a semiconductor wafer 202 is fixed on a wafer stage 201 provided in the prober. A probe card 204, having probe needles 203, . . . ,203 made of for example tungsten, is disposed above the semiconductor wafer 202. These probe needles 203, . . . ,203 are brought into contact with IC terminals on the semiconductor wafer 202, so that an electric power voltage or signal can be supplied to the integrated circuit by means of a tester or the like to detect an output signal from the integrated circuit chip by chip. For testing the same kind of integrated circuits in a short time, a full automatic prober is normally used since it has an alignment function and is capable of automatically executing a measurement of chip one by one. In FIG. 32, a reference numeral 205 represents a wiring pattern and a reference numeral 206 represents an external electrode terminal.
Hereinafter, a conventional testing method of a semiconductor wafer using a full automatic prober will be explained with reference to FIGS. 32 and 33.
First of all, in a step SB1, the semiconductor wafer 202 is automatically transported from a wafer carrier onto the wafer stage 201. Next, in a step SB2, positioning of the semiconductor wafer 202 is carried out using a CCD camera or the like so that IC terminals on the semiconductor wafer 202 can be brought into contact with the probe needles 203, . . . 203. Then, in a step SB3, the wafer stage 201 is shifted below the probe card 204 so that the semiconductor wafer 202 is placed below the probe card 204.
Subsequently, in a step SB4, the probe needles 203, . . . ,203 are brought into contact with the IC terminals on the semiconductor wafer 202. An electric power voltage or signal is applied to the integrated circuit to measure an output signal from the integrated circuit, thereby executing a test of the integrated circuit. After finishing the test of one integrated circuit, the wafer stage 201 is shifted to the next integrated circuit. Then, the probe needles 203, . . . ,203 are brought into contact with the terminals of the next integrated circuit to execute a measurement of the next integrated circuit.
According to the conventional testing method of a semiconductor wafer using a full automatic prober, a plurality of integrated circuits on the semiconductor wafer 202 are successively measured in the manner above-described. When the test of all the integrated circuits is completed, the semiconductor wafer 202 is returned from the wafer stage 201 to the wafer carrier in a step SB5. For a plurality of semiconductor wafers 202, . . . ,202, above-described steps are repeatedly executed to accomplish a measurement of each semiconductor wafer 202. When the measurement of all the semiconductor wafers 202 is finished, operation of the full automatic prober ends.
A method of shortening a test time per chip would be, for example, to provide a self test circuit (i.e. BIST circuit) to execute the burn-in screening (high-speed operation) of memories such as DRAM by means of a prober.
Executing the burn-in screening processing in the wafer condition according to the previously-described testing method of a semiconductor wafer using a prober would require a time not longer than 1 minute in total for the shifting of the semiconductor wafer 202 in the procedures of the steps SB1, SB3 and SB5 and the positioning of the semiconductor wafer 202 in the step SB2. However, the burn-in screening in the step SB4 usually requires several to several tens hours. The conventional testing method of a semiconductor wafer using a prober is disadvantageous in that it necessitates to test semiconductor wafers one by one. Accordingly, it takes an extremely long time to test a great amount of semiconductor wafers. This will results in a huge increase of cost for manufacturing an LSI chip.
Another disadvantage of the testing operation using an automatic prober is an exclusive usage of the probe during tests, because the alignment function cannot be used for tests for other kinds of semiconductor wafers or other purposes.
Providing a BIST circuit for shortening a test time per chip, applied to DRAM or the like, leads to an increase of a chip area and reduces the number of chips per wafer, thus causing a problem of increasing chip cost.
To execute the burn-in screening of bare chips at a time in the wafer condition, it is necessary to simultaneously apply an electric power voltage or signal to a plurality of chips formed on the same wafer, to operate all of these plural chips. To this end, it will be necessary to prepare a probe card having numerous probe needles (e.g. several thousands or more). However, the conventional needle type probe card cannot meet such a need in view of great number of pins and cost increase.
Proposed to solve such a problem is a thin film type probe card having bumps on a flexible substrate (Refer to Nitto Technical Reports Vol. 28, No. 2 (October 1990) PP. 57-62)
Hereinafter, the burn-in screening using a flexible substrate with bumps will be explained.
FIGS. 34(a) and 34(b) are cross-sectional views illustrating the probing condition when a flexible substrate with bumps is used. In FIGS. 34 (a) and 34(b), a reference numeral 211 represents a probe card which comprises a polyimide substrate 218, a wiring layer 217 formed on the polyimide substrate 218, bump electrodes 216, . . . ,216, and a through hole wiring connecting the wiring layer 217 and the bump electrodes 216, . . . ,216.
As illustrated in FIG. 34(a), the probe card 211 is pushed against a semiconductor wafer 212 serving as a tested substrate so that a pad 215 on the semiconductor wafer 212 is electrically connected to a corresponding bump 216 of the probe card 211. If testing condition is in a room temperature, a test will be feasible in this condition by simply applying an electric power voltage or signal to the bump 216 via the wiring layer 217.
However, a diameter of the semiconductor wafer 212 possibly increases up to, foe example, 6 inches in the probing using the conventional probe card 211, causing deflection of the semiconductor wafer 212 and/or unevenness in height of the bump 216 which lead to a first problem of making some of electrical connections useless between the bumps 216, . . . ,216 and the pads 215, . . . ,215.
The burn-in screening generally requires a step of increasing the temperature of the semiconductor wafer 212 to execute temperature acceleration. FIG. 34(b) shows a cross-sectional structure of the semiconductor wafer 212 heated from the room temperature 25.degree. C. to 125.degree. C. In FIG. 34(b), the left-hand portion shows condition of a center of the semiconductor wafer 212, while the right-hand portion shows condition of a periphery of the semiconductor wafer 212.
Polyimide constituting the polyimide substrate 218 has a thermal expansion coefficient larger than that of silicon constituting the semiconductor wafer 212. (More specifically, the thermal expansion coefficient of polyimide is 16.times.10.sup.-6 /.degree. C., while the thermal expansion coefficient of silicon is 3.5.times.10.sup.-6 /.degree. C.) Thus, dislocation between the bump 216 and its corresponding pad 215 is found at the peripheral portion of the semiconductor wafer 212. More specifically, if the semiconductor wafer 212 of 6 inches and the probe card 211 are aligned in position at a room temperature and then heated to 100.degree. C., the probe card 211 will cause a thermal expansion of 160 .mu.m while the semiconductor wafer 212 will cause a thermal expansion of 35 .mu.m. In other words, a dislocation between the pad 215 and the bump 216 will increase up to an approximately 125 .mu.m at the outermost periphery of the semiconductor wafer 212. Such a dislocation due to difference of thermal expansion is so serious that electrical connection between the pad 215 and its corresponding bump 216 cannot be maintained in the peripheral region of the semiconductor wafer 212.
As explained above, according to the conventional burn-in screening, the semiconductor wafer is heated during the burn-in screening. The probe card brought into contact with the semiconductor wafer is also heated. Thus, difference of thermal expansion coefficients between the semiconductor wafer and the probe card causes a serious dislocation therebetween, resulting in a second problem that electrical connection between a pad and its corresponding bump cannot be maintained at the periphery of the semiconductor wafer.
To execute the burn-in screening of bare chips in the wafer condition, it is necessary to operate each bare chip by simultaneously applying an electric power voltage or signal to a plurality of bare chips formed on a single semiconductor wafer. However, supplying an electric power voltage or signal to each bare chip independently is not practical in view of cost since the number of wiring patterns exclusively formed on the semiconductor wafer increases up to several thousands or several tens thousands. To reduce the number of independently or exclusively provided wiring patterns, it is necessary to commonly use the wiring patterns as many as possible.
However, if extraordinary current flows through one bare chip connected to such a common wiring pattern, adverse effect of extraordinary current will spread to other bare chips associated. Thus, it becomes impossible to execute an ordinary burn-in screening.
Accordingly, in executing the burn-in screening, it is necessary to electrically remove such an extraordinary bare chip from the common wiring pattern.
Hereinafter, a method disclosed in the Unexamined Japanese Patent Application No. HEI 1-227467/1989 will be explained as one example of the burn-in screening of bare chips in the wafer condition.
FIG. 35 shows one of plural bare chips formed on a semiconductor wafer. In FIG. 35, a reference numeral 240 represents a bare chip, a reference numeral 243 represents an electric power source pad of the bare chip 240, a reference numeral 244 represents a GND pad of the bare chip 240, a reference numeral 241 represents a burn-in electric power source pad, and a reference numeral 242 represents a P-channel type transistor. The transistor 242 has a drain connected to the burn-in electric power source pad 241 and a source connected to the electric power source pad 243. Furthermore, in FIG. 35, a reference numeral 245a represents a first pad connected to the gate of transistor 242, and reference numerals 245b and 245c represent second and third pads respectively connected to the first pad 245a via a thin aluminum pattern of, for example, 3 um width. A first resistance 246a is interposed between the GND pad 244 and the second pad 245b. This first resistance 246a has a relatively low resistant value (e.g. 10 k.OMEGA.). A second resistance 246b is interposed between the burn-in electric power source pad 241 and the third pad 245c. This second resistance 246b has a relatively high resistant value (e.g. 100 k.OMEGA.).
First of all, before a burn-in screening, a wafer test is executed by probing electric power source pad 243, GND pad 244, first to third pads 245a-245c using a stationary probe needle (not shown) connected to an external measurement device (not shown). The external measurement device applies an electric power voltage to the electric power source pad 243 and applies a GND voltage to the GND pad 244. Furthermore, the first pad 245a is given an "H" level signal. As the transistor 242 is in an OFF condition under such a condition, no current flows between the electric power source pad 243 and the burn-in pad 241. Thus, a test of the bare chip 240 is feasible.
When the bare chip 240 is found to be a non-defective as a result of the test of the bare chip 240, the next bare chip is subsequently tested. On the contrary, if the bare chip 240 is found to be a defective, the external measurement device supplies large current (e.g. 100 mA) between the first pad 245a and the second pad 245b to fuse the shin aluminum pattern interposed between the first pad 245a and the second pad 245b before going on a test of the next bare chip.
By doing the wafer test in this manner before the burn-in screening, the burn-in screening can be effectively carried out. More specifically, when the burn-in electric power source pad 241 is applied an electric voltage in the burn-in screening, the transistor 242 is turned on in the case the bare chip 240 is found a non-defective in the wafer test because the gate voltage of the transistor 242 becomes an "L" level due to existence of the second resistance 246b having a resistant value larger than that of the first resistance 246a. Thus, the burn-in voltage is supplied to the electric power source pad 243 through the transistor 242. On the other hand, the thin aluminum pattern between the first pad 245a and the second pad 245b is fused off in a case the bare chip 240 is found a defective in the wafer test. Therefore, the gate voltage of the transistor 242 becomes an "H" level, and the transistor 242 is turned off. Thus, the burn-in voltage is not supplied to the electric power source pad 243, preventing electric current from flowing through the defective bare chip 240.
As described above, electric power source current is surely prevented from flowing through the defective chip even if an electric voltage is applied to the burn-in electric power source pad 241. Hence, adverse effect is not given to the non-defective chips in the burn-in screening.
However, above-described arrangement requires to form on each bare chip 240 excessive elements such as transistor 242, first and second resistances 246a, 246b, burn-in electric power source pad 241, first to third pads 245a-245c and aluminum pattern acting as a fuse. Furthermore, an electric voltage is applied through the transistor 242 in the burn-in screening and, therefore, an electric voltage applied to the burn-in electric power source pad 241 is not directly applied to the internal electric power source. Thus, a third problem arises in that a significant voltage drop is induced in the bare chip 240.